how to deploy a CPU-only version of PyTorch
I am trying to deploy a FastAPI application in a Docker container and I think I need to deploy a CPU-only version since as I understand Railway doesn't support GPU yet.
Do anyone in here know how to deploy a CPU-only version of PyTorch and related libraries, including torchaudio?
I have seen that the Demucs repo has an "environment-cpu.yml" file: https://github.com/facebookresearch/demucs/blob/main/environment-cpu.yml
7 Replies
Project ID:
5ab1b9a6-5425-437f-a235-b68de753018eproject ID: 5ab1b9a6-5425-437f-a235-b68de753018e
wouldn't it run on the CPU by default if it can't find any GPUs?
also this doesn't seem like a Railway issue, help threads are specifically for Railway related issues
@Fragly I know this is not a Railway issue, but I was just wondering if it was possible to ask for help about this in here
I am trying to deploy to Railway at least
I understand, although wouldn't it be more fitting to ask this question in a PyTorch community?
@Fragly you're right, I will delete this post, do you happen to know any PyTorch communities?
well there's the PyTorch forums, I bet there's a discord server related to PyTorch too, will just need to search for it
Also no need to delete the thread, feel free to open a new thread if you ever have an issue or question about Railway 🙂